Meet SAM, the tutor that teaches back.
watch one session: it sees, asks, remembers, and acts
SAM reads what you write, asks the question that lets you find the slip yourself, updates what it knows about you from a real check, and plans when you should see it again.
Four faculties, one agent.
the anatomy of a mentor
What separates SAM from a search box: it doesn’t just answer. It perceives, remembers, reasons pedagogically, and acts — the loop a real tutor runs. You just watched all four in one session.
perception
It sees what you write
Vision reads your handwriting and sketches on the board, so SAM responds to your actual attempt — not a multiple-choice proxy.
memory
Assessment history persists
Supported formal assessments can update mastery estimates that remain available between sessions.
pedagogy
It knows how people learn
Bloom’s taxonomy, Bayesian knowledge tracing and spaced repetition decide what comes next — pedagogy, not guesswork.
agency
It acts on its own
SAM draws, narrates, schedules reviews and nudges you when it senses struggle — before you have to ask.
One loop, running continuously.
how SAM thinks
Supported assessment and review interactions can feed the same cycle, so your results shape what SAM recommends next, while keeping the scope of each stored result explicit.
Capabilities, not party tricks.
what SAM can do
The Board Tutor
Teaches on a living whiteboard — sketching diagrams, writing equations, and narrating each step out loud.
Mastery tracking
Supported formal assessments can update a Bayesian mastery estimate.
Spaced repetition
An SM-2 schedule resurfaces each concept right before you’d forget it.
Bloom’s progression
Work is graded by cognitive depth — Remember through Create — not just right or wrong.
Proactive nudges
SAM watches for struggle and offers help before you have to ask for it.
Depth-aware assessment
Exams in traditional and Socratic modes, with diagnostic, misconception-aware grading.
Built on frontier models, with a fallback if one blinks.
under the hood
SAM routes each task to the right model and falls back automatically if a provider fails. Vision-capable where it counts, so it can actually read what you draw on the board.
- Anthropic Claude
- Google Gemini
- OpenAI
- DeepSeek
- Mistral
- 30+
- educational engines
- 6
- Bloom's levels
- 4
- core faculties
Let SAM teach you something.
Pick one thing you want to understand. SAM handles the board, the memory, and the next step.